A potential $25 billion funding round, led by SoftBank, could value OpenAI at a staggering $300 billion, a move set against the backdrop of seismic shifts in the artificial intelligence sector.
The rise of cost-effective models from China’s DeepSeek, and the unraveling of OpenAI’s exclusive cloud partnership with Microsoft, are creating major industry shifts and strategic repositioning. This investment, along with the ongoing Stargate Project, seeks to solidify OpenAI’s position as a dominant force in the AI space.
SoftBank’s Direct Investment Plans
SoftBank’s commitment extends beyond infrastructure, with ongoing discussions for a direct equity stake of $15 to $25 billion in OpenAI, alongside their $18 billion commitment to the Stargate initiative, according to The Financial Times.
Simultaneously, OpenAI is expected to invest around $15 billion in the Stargate Project, partly funded by SoftBank’s investment. This signals a shift away from reliance on Microsoft for compute resources and marks the end of their exclusive cloud agreement. Microsoft, for its part, is committed to a large investment in it’s own AI infrastructure.
Masayoshi Son, SoftBank’s CEO, has acknowledged missed opportunities in AI, now focusing on Sam Altman, OpenAI’s CEO. Son reportedly wrote in an email to investors, “For all of us, the AI era represents a once-in-a-generation opportunity to help build a better, safer, healthier, and more prosperous future.”
This signals a strategic change for SoftBank following previous missteps, including their $16 billion investment in WeWork. The proposed deal would make SoftBank the largest backer of OpenAI, occurring as DeepSeek’s AI models are causing market unease, with a temporary dip in Nvidia’s stock, as investors question the need for expensive AI hardware.
The Stargate Project: Independence Through Infrastructure
The Stargate Project, a collaboration among OpenAI, SoftBank, and Oracle, aims to establish extensive data centers across the U.S., backed by the Trump administration. SoftBank’s CEO Masayoshi Son characterized it as “a strategic capability to protect the national security of America and its allies.”
With an initial investment of $100 billion, with a potential expansion up to $500 billion over four years, the project includes Microsoft as a technical partner, though a “right of first refusal” clause requires OpenAI to use Microsoft’s Azure cloud as priority if they meet their needs. The project also incorporates custom chip development with TSMC and Broadcom, with an expected release in 2026.

The funding will come through 20% equity and 80% from debt secured by assets and cash flow, along with participation from Oracle and the United Arab Emirates fund MGX. A data center already in Texas is set to be included within the initial $100 billion investment, but the project will need billions beyond initial commitments.
DeepSeek’s R1 Model is Challenging the Norms
DeepSeek’s R1 model has come to challenge the conventional approach, showing high performance with fewer resources, using 2,048 Nvidia H800 GPUs. This has raised questions about whether DeepSeek is actually using more powerful Nvidia H100 GPUs, despite export restrictions, as suggested by Scale AI CEO Alexandr Wang on CNBC.
The model and the question of its hardware has triggered debate about the cost-efficiency of AI development and the potential for more accessible models.
However, OpenAI has presented evidence that DeepSeek may have trained its models using a technique called distillation, by which a smaller model mimics the performance of a larger one. OpenAI believes this breaches their terms of service.

While OpenAI investigates, Microsoft has made DeepSeek R1 available on its Azure AI Foundry and GitHub Models, where it can be compared side by side with other AI models. Microsoft’s VP of AI Platform, Asha Sharma, stated, “One of the key advantages of using DeepSeek R1 or any other model on Azure AI Foundry is the speed at which developers can experiment, iterate, and integrate AI into their workflows.” This move has created further scrutiny surrounding AI ethics and data security.
OpenAI today is reportedly quickly releasing its o3-mini model today for immediate access via API and ChatGPT. This fast-tracked launch seeks to compete with DeepSeek by offering o1-level reasoning at 4o-level speeds.
The release, which is part of the o3 family, also includes the o3-mini-high model, both aimed at improving problem solving in areas like coding, science, and mathematics, while also offering a faster response time than previous models.
These developments were hinted at previously by OpenAI CEO Sam Altman and through updates found in the latest ChatGPT web build which has a pop-up introducing the models.
The new models come in two versions: o3-mini and o3-mini-high. Each is optimized for reasoning tasks, but vary in compute settings to give users flexibility in performance.
o3-mini and o3-mini-high reportedly outperform the o1 pro model in speed, and, in some scenarios match, or exceed its reasoning capabilities. The models handle complex queries with lower latency, which makes them suitable for both casual use, and for professional applications. Users will find the option to try out those models via a new pop-up in ChatGPT, and will be able to do so using the same user interface as the current o1 experience.
BREAKING 🚨: OpenAI is about to announce 2 new reasoning models soon: "o3-mini" and "o3-min-high".
— TestingCatalog News 🗞 (@testingcatalog) January 31, 2025
"o3-mini-hype" 👀👀👀 https://t.co/pXrxiiunZX pic.twitter.com/UtHHcpKhFC
The Microsoft-OpenAI Relationship: A Power Shift
The partnership between Microsoft and OpenAI is shifting as OpenAI seeks greater independence. While Microsoft CEO Satya Nadella has declared, “OpenAI APIs are exclusive to Azure going forward even, so nothing changes there… OpenAI has committed to Azure in a very significant way,” the Stargate project introduces the potential for using other providers.
Nadella admitted that their partnership was recently adjusted to allow OpenAI greater flexibility: “More importantly, I think for us, one of the things that—in fact, Stargate is the answer. Sam wants to continue with the scaling laws to build out more compute… we come to us first, but if we can’t meet those needs, he can go elsewhere.”

Nadella also reinforced Microsoft’s commitment, saying, “All I know is I’m good for my $80 billion. I am going to spend $80 billion building out Azure. Customers can count on Microsoft with OpenAI models being there everywhere.”
The AGI clause, that mandates that upon achieving artificial general intelligence, that control will be transferred to a non-profit board, is also under review by OpenAI which might restrict Microsoft’s access.
An Increasingly Competitive AI Market
The AI industry is now a much more competitive field with companies such as Google, Anthropic, and Elon Musk’s xAI advancing their models. Musk questioned the viability of Stargate and SoftBank’s funding, tweeting: “SoftBank has well under $10B secured. I have that on good authority.”
Altman responded to Musk’s comments, saying “want to come visit the first site? Its great for the country. maybe put America first”, resulting in President Trump dismissing Musk’s comments as a result of normal personal conflict.
DeepSeek’s Ethical Scrutiny
DeepSeek, the new emerging AI star from China, meanwhile faces scrutiny concerning its data practices and the spreading of misinformation, showing a pro-China bias. NewsGuard’s review has shown that DeepSeek returns inaccurate responses 83% of the time, often promoting false claims about topics like the Syrian conflict, or the 2024 Azerbaijan plane crash, also including a Chinese point of view. The U.S. Navy banned the use of DeepSeek’s AI models citing security concerns.
Additionally, Italy’s regulater Garante is investigating DeepSeek’s compliance with GDPR and data practices in China, while Microsoft and OpenAI investigate possible theft of data used in training their model. Investigative journalist George Webb, has linked DeepSeek to the death of a former OpenAI researcher, Suchir Balaji.
Microsoft’s Copilot Offers Free Access to o1
To gain a more competitive position, Microsoft is now offering access to OpenAI’s o1 model for free to all Copilot users, removing the requirement for a paid Copilot Pro subscription for the “Think Deeper” feature, which employs o1 for multi-step reasoning. Microsoft A I CEO Mustafa Suleyman said, “we’ve made Think Deeper free… now everyone gets OpenAI’s o1,” also hinting at future expansions. This positions Copilot as a more appealing alternative than OpenAI’s subscription model, by providing its advanced features for free.